While Heidi is introducing AI agents capable of performing complex clinical admin tasks and Amplar Health is testing automated patient risk identification, Aspen Medical says it is not yet convinced the benefits of agentic AI outweigh the risks.
Australian healthcare companies are taking markedly different approaches to deploying agentic AI, with some introducing systems capable of performing complex tasks while others remain cautious about the risks of greater autonomy.
Speaking at The Medical Republic and Health Services Daily’s AI and Hospitals summit last week, Aspen Medical chief medical officer Dr Katrina Sanders said her organisation had not deployed agentic AI.
“We have not deployed any agentic agents. That is not the healthcare company we are,” she said.
“We’re highly practical, and we’re not convinced about the risk-reward benefit at this point in time.”
Instead, she said Aspen was concentrating on improving its underlying data infrastructure and conducting AI pilots to assess potential returns on investment and scalability.
This article originally ran on TMR’s sister site, Health Services Daily. TMR readers can sign up for a discounted subscription.
Dr Sanders said the company was dealing with 902 systems and 40,000 SharePoint files as part of preparations for future agentic AI applications.
She identified operational activities, including workforce deployment, logistics and business administration, as the areas where Aspen expected the greatest benefits.
Heidi co-founder Dr Tom Kelly described how the company’s recently introduced agentic capabilities could move beyond clinical documentation to undertake sequences of administrative tasks.
Using hospital discharge as an example, Dr Kelly said AI agents could review patient records, identify clinical problems, consult guidelines, prepare discharge instructions and determine which healthcare professionals needed to receive information.
“If something is a task that is mostly in the digital world or mostly in software, you can now set up agents to do that work without a human sitting there driving the keyboard,” he said.
However, Dr Kelly stressed that clinical decisions remained subject to human oversight.
“If anything is making diagnosis, informing investigation or management, it has to be reviewed and approved by clinician,” he said.
Amplar Health’s clinical design manager Rob Edwards outlined a more cautious application, with the company developing AI to review patient-reported outcome measures and identify emerging clinical risks.
Related
Mr Edwards described a scenario in which a patient’s overall falls-risk score was improving, but their latest responses revealed they had begun forgetting to take medication.
The AI could identify the discrepancy and flag it for a clinician.
“It’s not making any decisions. It’s not making any determinations. It’s recommendations only,” Mr Edwards said.
He said Amplar was deliberately taking a risk-averse approach, with clinicians triggering reviews, assessing the outputs and providing feedback before any move towards greater autonomy.
Meanwhile, Dr Kelly warned that healthcare organisations could struggle to establish appropriate governance as AI systems became increasingly capable.
“I think the capability will outstrip like our ability to put it to work in a way that we can govern, basically,” he said.
He argued that technology companies should accept liability for their products and that AI agents should operate within auditable environments, using their own system credentials rather than clinicians’ logins.
Mr Edwards also emphasised the importance of workforce training and involving clinicians in designing systems that addressed genuine problems.
Dr Sanders said effective AI governance required established clinical safety principles, including quality assurance, auditing, monitoring and organisational oversight.
The TMR and HSD summit was held in Melbourne on Thursday 29 October 2026.
